Professional Documents
Culture Documents
White Paper
dB
+4
+1
-1
-6
- 30
(147 bits)
t
10 Ps 8 Ps 10 Ps 542.8 Ps 10 Ps 8 Ps 10 Ps
Time mask for normal duration bursts at GMSK modulation.
3
The output of the TxVCO will typically have a tolerance A relevant characteristic of the ACPM-7891 is the dynamic
of ±2 dB with also typical nominal output power level of range of approx 0.8 V between 0 to 35 dBm of output
around 5 dBm. At this stage the RF carrier with the modu- power level. Any control loop system will require a high
lation information is ready to be amplified by the PA. With level of discrimination in Vapc in order to accurately set
a PA input power level of around 0 dBm, the power control the correct output power level. This is a challenge for the
loop requirement is to control that power level following design of the power control loop as a few millivolts may
the GSM specified power level steps from output power well represent a change of power level. Furthermore, at
of +5 to +33 dBm. low power levels the discrimination of Schottky diode
power detectors is at its lowest, making the resolution
Voltage controlled PA even smaller.
In order to illustrate the response of a GSM PA, we have
plotted the actual response of the Avago ACPM-7891 Control loop theory
Tri-Band PA for the GSM, DCS and PCS bands versus control
Automatic gain control (AGC)
voltage. The input to the PA would be a GMSK modulated
RF carrier of constant power level of 0 dBm and the PA AGC is widely used in communication systems to maintain
maximum output level is of around 35 dBm. Input RF constant signal strength. As mentioned previously,
carrier and Vapc are both pulsed following the GSM TDMA changes in performance of each individual PA, tolerances
characteristic response. This is a period of 4.615 ms with in different components of the transmitter chain, supply
a duty cycle of 12.5% for standard GSM (1/8). The graphs voltage variations and changes in performance across
clearly reflect the characteristic output power with “output frequency, are all intrinsic to the system at nominal tem-
power response of the PA against voltage control with a perature. Some of them will be taken into account at the
steeper slope at lower power levels and flat gain response phone alignment stage of the manufacturing process.
when the PA gets into saturation. However, the main purpose of the alignment is to set a
look up table voltage values corresponding to different
ACPM-7891 response controlled power levels. Generally changes of power level
40
35 across supply voltage and frequency range will not be cal-
30 ibrated in the phone alignment during the manufacturing
25
OUTPUT POWER (dBm)
Modulator
Q(t)
Q(t)cos (2Sft + I)
Phase detector Loop filter Tx VCO PA
S/2 Tx IF Tx RF
6 I
BPF
Tx IF
I(t) I(t)sin (2Sft + I)
sin (2Sft + I)
/N /P Vcontrol
Offset mixer
VCO
4
Gain Control loop theory
A typical AGC loop is a feedback system comprising a The design of E, in addition of the linear gain required to
forward gain stage (A), feedback gain (E) and a signal optimize the response of the different components used
comparison stage that generates a differential error in the control loop (GSM systems will require an attenu-
signal. AGC loop is analyzed in terms of its closed loop ator, comparator and reference voltage source) might
gain (forward transfer function) and open loop gain. R(s) require an integrator depending on the loop type. Each
is the input amplitude and C(s) represents the output integrator within the loop will add a pole:
amplitude.
In the frequency domain:
Vref Vs(s) 1 1
R(s) E(s) C(s)
= = ;
Ve(s) 1+ RCs 1+ RCjZ
6 A
(+) j
( ) with its pole at Z =
RC
In control theory, the number of poles of the transfer
B(s)
E function is what determines its type. Poles are values
of ‘s’ (jZ) that make the denominator of the closed loop
transfer function equal to ‘0’ (notice that this would be the
Control loop relevant equations are: same as making the open loop transfer function equal to
Close loop gain: C(s)/R(s) = A/(1+AE) ‘-1’). The loop Type refers to the order of the open loop
transfer function pole. The number of poles required in
Open loop gain: B(s)/R(s) = AE
the open loop transfer function (AE) to obtain constant
Characteristic equation: 1/(1+AE) output signal will be determined by the way the output
Response of an AGC loop to the system output amplitude signal amplitude changes, it can generally follow a step,
fluctuations –changes in C(s) value- depends on its closed ramp or parabolic function.
loop transfer function as the R(s) reference signal will be A loop with an amplitude variation following a step function
fixed (fix characterized value at every output power level is of type 0 and needs no integrator in the open loop
of the mobile phone system). Variations in the forward gain transfer function. Amplitude variations following a ramp
value (A) due to voltage supply, operating temperature or function will characterize a loop type 1 which needs one in-
drive reasons, is what will originate those amplitude fluc- tegrator in the open loop transfer function and amplitude
tuations at the output C(s). Control loop feedback gain (E) variations following a parabolic function will characterize a
has to be designed to respond to those amplitude fluc- loop type 2 which would need two integrators.
tuations and correct them in order to obtain a constant
steady state output signal C(s). It is important to note that with a non-continuous (due to
the characteristic TDMA time multiplexing) fast response
The question for the loop designer is to obtain a model of control loop, all changes in amplitude in the control loop
the response of A in order to determine E so the system input can be regarded as instantaneous, hencethe control
keeps C(s) constant while meeting control loop stability loop is of type 0. In such cases, no integrator is required in
criteria (those criteria are discussed later). the feedback transfer function (E).
All real amplifiers have a number of internally compen- The loop design exercise will then be mainly about the
sated poles so that they can be represented as having a implementation of the comparator. Once the transfer
single pole close above the higher operating frequencies. function of the comparator is determined, the adequate
The power amplifier in the loop can therefore be modeled op-amp implementation can be selected to meet the
as having a transfer function with variable gain and a characteristic loop gain and speed requirements.
dominant pole at frequencies above those operating fre-
quencies. This would be somewhere above 1 GHz in the
case of a GSM power amplifier.
Typical PA gain in the frequency domain:
Vs(Z) A(Vapc)jZ
= with its pole at Z = Za.
Ve (Z) Z ;
1+ j
Za
5
Basic stability theory
Stability theory will be discussed only at a very elemen- Stability requires PM > 0 and GM > 1. It is generally accepted
tary level. It is not the purpose of this paper to take this that any system needs to be designed in principle with
complex issue into any real depth. However some basic a minimum GM = 10 dB and PM = 45 degrees. In Type 1
principles are highlighted below. control loops, the presence of two poles in the transfer
function with no zeros compensating each pole phase
Control theory uses the root locus plane to represent the
shift, will add two S/2 degrees of phase shift. At certain
zeros and poles of the open loop transfer function AE: it
frequencies the PM will therefore be ±S and make the
can be represented as module and phase components AE
system unstable. It has to be remembered that although
=(V + jZ). Vector positions across the frequency spectrum
the mobile phone control loop is of Type 0, the presence
will determine whether the system is potentially unstable
of a LPF to limit the loop BW will add an integrator to the
at any frequency. The control loop system will be unstable
system. Stability needs therefore to be monitored as the
when the following conditions are met:
power amplifier transfer function may have phase shift
_V_ > 1
I#S }
; I = arctn (Z/V) across some specific frequencies that could potentially
cause instability.
6
Loop gain and bandwidth Control loop components
The objective of the loop is to compensate for any varia- The required dynamic range of the control loop, wide
tions at the Power Amplifier (A) due to changes in perfor- non-linearity of the response at block level (PA and power
mance, temperature or voltage supply under the presence detector) and variation with temperature of Vf at the diode
of a fixed voltage reference level ensuring that the output detector, make the control loop design an interesting
power level C(s) is constant. This has to be achieved for a challenge. Three elements need to be correctly specified:
number of power steps or fixed Vref levels. The non-linear
PA and Detector Diode response causes a variation of the x Diode detector
close loop gain C(s)/R(s) at different output power levels: To maximize its dynamic range to cover the input power
from a factor of 1 for output power level of 5 dBm to a level, proper selection of its biasing conditions and input
factor of 6 for output power level of 33 dBm. Those non- power level needs to be made. Selection of the best-suited
linearities of detector and power amplifier dictate a high coupler-coupling factor and adequate RF 50 : termina-
level of discrimination at Vref for high power levels and tion needs to be provided at the diode input. The output
a very low level of discrimination for lower power levels. load should guarantee the diode optimum functionality
The accuracy of the Vref voltage control source (DAC on by setting its through current. RF de-coupling capacitors
the mobile phone controller) needs to be enough for the at the diode output will remove any RF signal and its
system to meet the tolerance requirements in GSM output harmonics.
power level specifications. The GSM standards have taken
x Temperature compensation
this into account and the tolerance at low power levels in
extreme conditions is approximately ±6 dB. There is a strong dependency with temperature of the
detector diode junction resistance (Rj) and consequently
We have previously seen that another requirement of with the forward voltage drop (Vf ). This difference of
GSM systems is for the carrier to meet a certain time mask. voltage drop with temperature may add up to the rectified
The rising and falling envelope of any radio frequency RF signal with the consequent detection of the wrong
carrier will generate a transient spurious response. power level. The solution is to use an identical diode
Those need to be kept under certain limits, hence it is working in the same way as the one used for RF detection,
necessary to ‘shape’ those profiles in order to minimize and it should be biased in an identical way. Temperature
the spurious emissions. The Vout reference DAC output compensation of the offset term (RF detected) is obtained
splits the 28 Ps of allocated time for the profiles into if the bias current of each diode is equal. Different com-
a number of amplitude ‘towers’ or registers. Common pensation schemes circuits can be used with the diodes,
number of registers is sixteen, lasting for 1.75 μs each. but the basic concept for compensation is to use the
Therefore the system will require a BW of approximate- identical Vf variation in temperature of one diode to com-
ly twice the sampling speed of 571 kHz. Loop BW of pensate for the variation in the other diode.
1–1.2 MHz are commonly accepted. This needs to be
taken into account by the loop design because there is a Two different temperature compensation schemes have
need for limiting the loop BW and consequently reducing been used experimentally in an Avago control loop dem-
feedback signals to the PA control input. A loop filter will onstration board. One control loop uses a differential
therefore be added to the system. model (900 MHz control loop) and the other one use a
feedback model (1800 MHz control loop). Both tempera-
ture compensation schemes worked well maintaining the
maximum variation from nominal condition below ±1 dB
at maximum power levels and meeting also the extreme
Comparator Power amplifier condition specifications at power level 17–19.
Vref 40
ACPM-7891 response
35
R(s) E(s) 30
C(s)
Output Power (dBm)
6
25
20
C 15
10
5
0
-5
(+)
-10
-15
-20
(- ) Α 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Vcontrol (V)
Vout
B(s)
T
E Pinput
7
x Comparator stage SAMPLE_#1
Frequency = 000 MHz
The desired response of the comparator can be easily Nominal Vbatt = 3.6 V
understood by using a graphic example. For a given Vref Feedback temperature compensation design
and three possible Comparator Voltage input signals, the
Pout (dBm)
output wanted signals should correspond one to each Max dB
Temp Vbatt: 3 V Vbatt: 3.6 V Vbatt: 4.2 V variation
other as per the diagram:
33.1 34 34.22 0.9
+55 25 25.15 25.24 0.15
9 9.7 10.2 0.7
24.8 24.94 25.01 0.14
+25 33.22 34.01 34.18 0.79
7.8 7.4 9 1.6
33.36 33.97 34.09 0.61
-20 24.3 24.44 24.52 0.14
3.5 4.7 5.3 1.2
For product information and a complete list of distributors, please go to our web site: www.avagotech.com
Avago, Avago Technologies, and the A logo are trademarks of Avago Technologies in the United States and other countries.
Data subject to change. Copyright © 2005-2010 Avago Technologies. All rights reserved.
AV02-2414EN - March 25, 2010